Another Inverter build

[Hedges]

Thanks for the good reference materials.
Some members at the backshed forum shared their design of inductor core saturation tester, but I did not have the time to build it and try.

Thanks for the mention of Backshed. It didn't turn up, just EEVblog, when I was searching from work.

Variac is good to drive transformer winding. Current shoots up fast, so series resistor makes it more controllable. Can also provide current sense, if you don't have a current probe. Current transformer can also be used, but has about 100A range for 333 mV output, so not much signal at 1A or so.

Variac output is ground referenced and scope BNC terminals are grounded. An isolation transformer on output of Variac lets you establish a new ground, and can step up or down the voltage as needed. I use a step-down for chokes that saturate at a couple volts, and more of the Variac range can be used.

If using ground-referenced voltage probes for both winding voltage and sense resistor, of course same node for both grounds. Can also use identical second winding of transformer under test for voltage sense.

My earlier models just used voltage or inductance ratio to determine turns ratio. Inductance measured with network analyzer, impedance analyzer, or voltage/current sensed by scope at line voltage. Inductance with second winding open/short gives coupling coefficient.

The Variac, saturation, BH curve approach allows extraction of parameters for Chan model (easier to do graphically than number crunching a million data points, at least when I've tried to program it.) Chan does not include leakage inductance or resistance, so I add those as separate elements determined by my earlier method.

I've also measured inrush applying DC voltage, repeat same polarity and alternate. First time of a given polarity it holds off current for time period longer than one AC phase before shooting up. Second time, re-enters saturation in a couple milliseconds. Reverse and it holds off longer for one time. The current transformer (which has a core) couldn't capture rise time and peak, but Fluke probe which is flexible Rogowski coil (no core) has 20 kHz bandwidth and could.

I've used the Chan model primarily for chokes, got better simulation of their performance in a circuit. Core materials differ, and vendors usually don't provide all the parameters.

 

[tinyt]

Thanks for the mention of Backshed. It didn't turn up, just EEVblog, when I was searching from work.

Variac is good to drive transformer winding. Current shoots up fast, so series resistor makes it more controllable. Can also provide current sense, if you don't have a current probe. Current transformer can also be used, but has about 100A range for 333 mV output, so not much signal at 1A or so.

Variac output is ground referenced and scope BNC terminals are grounded. An isolation transformer on output of Variac lets you establish a new ground, and can step up or down the voltage as needed. I use a step-down for chokes that saturate at a couple volts, and more of the Variac range can be used.

If using ground-referenced voltage probes for both winding voltage and sense resistor, of course same node for both grounds. Can also use identical second winding of transformer under test for voltage sense.

My earlier models just used voltage or inductance ratio to determine turns ratio. Inductance measured with network analyzer, impedance analyzer, or voltage/current sensed by scope at line voltage. Inductance with second winding open/short gives coupling coefficient.

The Variac, saturation, BH curve approach allows extraction of parameters for Chan model (easier to do graphically than number crunching a million data points, at least when I've tried to program it.) Chan does not include leakage inductance or resistance, so I add those as separate elements determined by my earlier method.

I've also measured inrush applying DC voltage, repeat same polarity and alternate. First time of a given polarity it holds off current for time period longer than one AC phase before shooting up. Second time, re-enters saturation in a couple milliseconds. Reverse and it holds off longer for one time. The current transformer (which has a core) couldn't capture rise time and peak, but Fluke probe which is flexible Rogowski coil (no core) has 20 kHz bandwidth and could.

I've used the Chan model primarily for chokes, got better simulation of their performance in a circuit. Core materials differ, and vendors usually don't provide all the parameters.

Pardon my ignorance. What is Chan model?

 

[Hedges]

Looking at the attached magnet wire data chart, the two strands #13 of the high voltage winding will have a cross section of 2 x 2.62 = 5.24 mm2 and the #10 low voltage winding has a cross section of 5.26 mm2.
Calculated HV winding current at 5KVA load is 5000 / 240 = 20.8A and each of the LV winding has a given full load rating of 31.25A.
Calculated HV winding current density is 20.8 / 5.24 = 3.97A per mm2.
Calculated LV winding current density is 31.25 / 5.26 = 5.94A per mm2.

Am I correct in guessing that they were able to rate the LV winding with high current because the winding uses short magnet wire and also on the outside which will have better heat dissipation?

I haven't tried to analyze heat in a transformer. I would start considering "enamel" (polyester?) thickness and number of turns.

You're guessing LV winding, secondary in the original application, is wound on the outside.
I'm not sure about these toroids, but power transformers are normally wound with primary on the outside (if one winding is over the other.)
Something about greater leakage inductance, and inrush is considerably higher if you drive them in reverse, applying voltage to secondary wound closer to core.
Worse than just inrush, my attempt to drive 480 Delta step-down to 120/208Y in reverse had excessive current draw and growled with 120V applied. Only down around 60V was it well behaved.

The toroids, used where efficiency matters, may also be sized and wound in a manner which keeps them further from saturation.

But you've disassembled far enough to confirm 10 awg secondary is on the outside. Maybe running that in reverse will perform better.

 

[tinyt]

I haven't tried to analyze heat in a transformer. I would start considering "enamel" (polyester?) thickness and number of turns.

You're guessing LV winding, secondary in the original application, is wound on the outside.
I'm not sure about these toroids, but power transformers are normally wound with primary on the outside (if one winding is over the other.)
Something about greater leakage inductance, and inrush is considerably higher if you drive them in reverse, applying voltage to secondary wound closer to core.
Worse than just inrush, my attempt to drive 480 Delta step-down to 120/208Y in reverse had excessive current draw and growled with 120V applied. Only down around 60V was it well behaved.

The toroids, used where efficiency matters, may also be sized and wound in a manner which keeps them further from saturation.

But you've disassembled far enough to confirm 10 awg secondary is on the outside. Maybe running that in reverse will perform better.

The pictures in posts #8 and #9 show that the LV winding (awg #10) is on the outside. These pictures were after I un-wrapped the outside mylar insulating tape.

To minimize work and materials, I intend to keep the existing inner HV windings in series with some of the existing LV windings as the secondary (output) and add the 26V primary (driven by the mosfets) as the primary in the inverter application.

 

[Hedges]

Pardon my ignorance. What is Chan model?

If you follow the links to forum threads I posted, one contains EEVblog:

Arbitrary (saturable) coupled inductors in LTSpice - Page 1

Arbitrary (saturable) coupled inductors in LTSpice - Page 1

eevblog.com

It explains the process and has downloadable LTSpice model. I highly recommend trying it. Free! Then by adding additional coupled windings you can model transformers and chokes.

That uses circuit elements to integrate voltage and get magnetization, same as a linked video from MIT.
The elements have to be tuned to frequency. What is better is to get the syntax for integration with math in LTSpice.

"Chan" was some student who developed a saturable, hysteretic inductor model in LTSpice.
Once you've got a BH curve (magnetization = integral of voltage in Y axis vs. current in X axis),
half width along X axis is Hc "Coercivity", magnetization current, the current required to bring magnetization back to zero (crosses X axis.)
half height along Y axis is Br "Remanence Magnetization", the magnetization which remains after returning from saturation to zero current
Projection of asymptotic tail of curve back to Y axis is Bs "saturation magnetization". The tail continues upward because the air core component never saturates, so projection gives the core's contribution.

Those three numbers are put into Chan SPICE model, which has simulated closer to measurements for me.

Because the curve is defined by just three numbers, I'm sure there are other effects in cores it doesn't capture. I would think different shapes would have corners getting different field strengths, entering saturation at different times. But the model is described with dimensions of a toroid. I've done some tortured adaptations of it to other shapes.

Another interesting exercise was trying to get core properties of a solid, not laminated, core. It was for a magnetic lens that uses DC to focus electrons. I could not get a saturation, just an oval BH curve with no tail. I applied 240Vrms to it and dissipated quite a few watts, about 300W or so as I remember. I "unrolled" the toroid into a bar shape in my calculations, determined resistance, treated it as a shorted winding, calculated that 2/3 of the power was dissipated in core and 1/3 in winding resistance.

Although I was not able to build a model I trusted, I was able to simulate magnetization and demagnetization with various dampened waveforms. No idea how accurate. I've dabbled with measuring BH curve by biasing with DC current in one winding and sensing inductance with another, but haven't completed that.

 

[Hedges]

To minimize work and materials, I intend to keep the existing inner HV windings in series with some of the existing LV windings as the secondary (output) and add the 26V primary (driven by the mosfets) as the primary in the inverter application.

It would be a shame to waste the existing wound secondary. Others (brandnewb) who gutted a transformer had regrets.
Can you use the 20V windings, or two in series 40V? If you use some windings in series with primary to change their turns ratio, might not be too far off.
What is your architecture?

I think modern "low frequency" inverters use inductors in buck configuration switched at high frequency to synthesize sine wave applied to transformer low-voltage primary windings. I think minimum battery voltage is related to transformer turns ratio.


The old Trace inverters were actually low frequency, apparently a MSW architecture where primary was driven directly to battery voltage, but multiple taps so MSW made a staircase approximating sine wave.
Seems to me the direct drive of a transformer allows massive current into low impedance, while HF buck converter is limited by what the core can store. Unless it is allowed to saturate, so inverter is MSW while driving surge current.

 

[Hedges]

The Chan model has an inductor, instead of L=<henries>, "Hc={Hc} Bs={Bs} Br={Br} A={A} Lm={Lm} Lg={Lg} N={1}
That inductor is modeled as 1 turn, and actual turns of each winding is referenced by a symbol which looks like an inductor, is actually a subcircuit that calls L1.
A is cross-sectional area, Lm is length of centerline of core, Lg is airgap (zero in the one I did.)

To plot magnetization of core, I calculated with a SPICE voltage source with value "V=idt(V(pin1)-V(pin4))" to get Webers, where "pin1" and "pin4" were terminals of the subcircuit. I had trouble finding that syntax until someone sent me a simulation that compiled successfully.

That was plotted vs. H in amps (maybe incorrect?) calculated by a SPICE voltage source V=I(L1)/N, so it accounted for N turns in a coil.

Alternate units, Tesla, I think, I ploted V=idt(V(coreB))/(A*N) vs. V=N*I(L1)/Lm

I never got my extracted units to agree with core vendor specs, maybe some scaling wrong, maybe not driven far enough into saturation (my extracted parameters varied.) But I did get simulation of noise attenuation by choke to show reasonably close to absolute and relative measurements in a test fixture.

 

[tinyt]

It would be a shame to waste the existing wound secondary. Others (brandnewb) who gutted a transformer had regrets.
Can you use the 20V windings, or two in series 40V? If you use some windings in series with primary to change their turns ratio, might not be too far off.
What is your architecture?

I think modern "low frequency" inverters use inductors in buck configuration switched at high frequency to synthesize sine wave applied to transformer low-voltage primary windings. I think minimum battery voltage is related to transformer turns ratio.


The old Trace inverters were actually low frequency, apparently a MSW architecture where primary was driven directly to battery voltage, but multiple taps so MSW made a staircase approximating sine wave.
Seems to me the direct drive of a transformer allows massive current into low impedance, while HF buck converter is limited by what the core can store. Unless it is allowed to saturate, so inverter is MSW while driving surge current.

For my reference when I modify the second toroid I made a drawing of the modficication stages (attached).
The four 20v windings are removed to make space in the center of the toroid for the new 26v winding.
I wanted to increase the current handling of all HV and LV windings together with the lowered core flux density. I have leftover AWG 14 magnet wires and I will be using it.
So, effective HV (now secondary) cross section will be from 5.24 mm2 to 5.24 + 2.08 = 7.32 mm2.
And the new 26V (now primary) outside winding cross section is 2.08 x 20 = 41.6 mm2. Using the earlier LV winding calculated current density of 5.94A/mm2, this 26V winding should be able to handle 41.6 x 5.94 = 247A (wow! , I don't believe it).

 

[tinyt]

Update:
I will be using this amber tape, green polyester tape, and 3M masking tape from home depot, also some polyurethane varnish.
I am also using wrapping tape salvaged from another toroid that I modified in 2018.

I have also added a 12v winding for sensing toroid voltage, an 8.5v winding to power the fans, and a 10kohms thermistor for temperature sensing. I have update my drawing for these.

 

[Warpspeed]

Looking at the attached magnet wire data chart, the two strands #13 of the high voltage winding will have a cross section of 2 x 2.62 = 5.24 mm2 and the #10 low voltage winding has a cross section of 5.26 mm2.
Calculated HV winding current at 5KVA load is 5000 / 240 = 20.8A and each of the LV winding has a given full load rating of 31.25A.
Calculated HV winding current density is 20.8 / 5.24 = 3.97A per mm2.
Calculated LV winding current density is 31.25 / 5.26 = 5.94A per mm2.

Am I correct in guessing that they were able to rate the LV winding with high current because the winding uses short magnet wire and also on the outside which will have better heat dissipation?

The recommended design parameters for a good cool running low idle current inverter transformer are a core flux density of 1 Tesla, and a current density in the wire of 4 amps per square millimeter. Those are very conservative numbers by commercial standards, most off the shelf transformers are way higher and will not work as well in an inverter.

Your measured idle current performance is so good, you need not worry about core saturation, or any input current surge at turn on.

 

[tinyt]

Update.
Finished the two additional awg 14 120v windings.

Also, after making sure phase directions are correct, soldered the jumpers and lead wires. Those awg 10 magnet wires (20v windings) are stiff. Will slobber some more polyurethane varnish for moisture protection and then hide the ugly soldering job with heatshrink tubing. I hope the solder connection will not fail.

 

[tinyt]

After winding another layer of wrapping tape, I started the added 26v winding. It is here that I have added the forgotten fan power winding. last photo shows the 26v 20 strands awg14 windings completed.

 

[tinyt]

I discovered that a 1.5 inch diameter medicine bottle makes a tight fit in the center of the toroid, so I used it to hold the wire strands in place in the center. Then I straightened the the winding ends and stripped them of insulation using this handy magnet wire stripper.

 

[tinyt]

I tested one of the 120V winding to measure current draw and it did not change, still 120 x .094 = 11.3 VAR. So looks like my 120v winding turns and jumper connections are all good, I should have done this right after winding the added 120v sections and also after the jumper connections, I guess am lucky. I used a variac to gradually increase the applied voltage to 120vac to avoid surge at turn on. In the absence of a variac a series incandescent lamp can be used.

 

[tinyt]

I wanted to make sure that each strand of the 26v winding has the same number of turns. So using a rubber band I tied each strand of the stripped start end together with a rubber band and with the toroid powered, I measured the output of each strand (other dmm probe at the finish end of each strand).

Eighteen measured 26.6vac and two measured 28vac. Looks like I made a mistake in counting the number of turns for these two. I am glad I did this, otherwise it will be a disaster. Luckily pulling one turn off from the two strands was easy.

 

[tinyt]

This time I trimmed the 26v wire ends so that the tips and stripped part will be even with each other when inserted in the crimp terminal, picture shows the top group is done. I also used heat shrink tubing and harness loom to make them nice.

 

[tinyt]

I used this crimping tool and these crimp terminals.

Pictures 35 and 26 show close up of the wires after crimping, I hope I applied enough pressure to make gas tight cold welded crimp.

If the crimp is gas tight at the cold weld, can I coat the exposed wire coppers to protect them from humidity and still keep the cold welded crimp intact?

 

[tinyt]

This part of the build is complete. The toroid weighs 42 lbs (19 kg).

I bought earlier the 48V 10000VA 220V/60Hz version of this power inverter. Will be checking it out next. I will probably need multiple 24 pcs HY4008 mosfets to prepare for fireworks. And also a fire extinguisher, LOL.

 

[tinyt]

Update:
The power inverter I bought does not use the popular EGS002 spwm module. It uses something marked "Century inverter SPWM DRIVE".

I cannot find the data sheet in the internet and the seller will not give it to me. He just describes the meaning of the flashing of the led. It does not have jumpers to change operating parameters. It uses STM8S003F3 instead of EG8010. There are solder pads 0 - 4. Maybe the un-used ones are used to change parameters. But there is no documentation.

Fortunately, the connector pin signal name matches those of the EGS002 which is fully documented. So, I ordered some and I will be using it instead.

 

[toms]

Very cool project! I always admire projects like this, I wish I could have afforded to DIY more components of my system.

BMS, Inverter, Chargers take too long (ie too expensive) to develop and build compared to what is commercially available.